fix: Force use index-based column mapping for old ORC file

[ccat3z]

An ORC file written by an old version has no field names in the physical schema, always map column names by index.

[netlify]

✅ Deploy Preview for meta-velox canceled.

Name	Link
🔨 Latest commit	18478bc
🔍 Latest deploy log	https://app.netlify.com/sites/meta-velox/deploys/67c17556239e44000844be93

[Yuhta]

This is hacky, your metadata service should keep this information and set useColumnNamesForColumnMapping=false accordingly for these files.

[ccat3z]

cc @kecookier

[ccat3z]

This is hacky, your metadata service should keep this information and set useColumnNamesForColumnMapping=false accordingly for these files.

It's impossible to check if it's old ORC files without reading it. And it's a logical error to use useColumnNamesForColumnMapping on these ORC files without column names. Spark will determine whether it is an old version when actually reading the ORC, perhaps an option like autoUseColumnPositionForOldOrc can be added to support this behavior?